Khushbu Shah

1 month ago | medium.com | Khushbu Shah

A great data pipeline isn’t just about moving data from one place to another; it’s about creating a system that’s reliable, easy to understand, and built to last. Real-time or batch, every pipeline is built to serve one purpose: getting reliable data into the hands of those who need it. But here’s the thing: data pipelines that inspire confidence don’t happen by accident.

1 month ago | medium.com | Khushbu Shah

You open your laptop on a regular workday. There’s a new request, build a data pipeline that syncs customer data, runs validation, and delivers analytics-ready tables. Normally, you’d start sketching out the DAG, writing boilerplate code, testing transformations, and tuning partitions. But not today. Today, you write a short prompt in Windsurf, and it generates the entire DAG for you. Meanwhile, Cursor has already written the SQL transformations, which are complete with tests and docstrings.

1 month ago | medium.com | Khushbu Shah

ProjectPro helps data professionals build job-ready skills through end-to-end solved projects in AI, Data Science, Machine Learning, and Data Engineering designed by industry experts. Learn by doing and accelerate your career with real-world hands-on experience. Follow publicationIf you’re building AI projects in 2025, you’ve probably realized by now, just knowing how to call an LLM API isn’t enough. A flood of AI tools, libraries, and frameworks promise to “10x” your development.

1 month ago | medium.com | Khushbu Shah

“RAG is dead.” That’s the hot take bouncing around the AI community right now, especially with all the buzz around models boasting 10 million token context windows. But let me stop you right there: this thinking is dangerously wrong. Yes, we now have models that can handle enormous amounts of text at once. Yes, 10M token context windows are impressive. But no, this doesn’t mean you should start dumping your entire knowledge base into every query and call it innovation.

1 month ago | medium.com | Khushbu Shah

“What if your AI could pull the perfect answer from the world’s knowledge every single time?”Retrieval-augmented generation (RAG) is the quiet revolution behind today's best-performing AI systems. ChatGPT’s ability to cite sources and enterprise AI’s scanning of thousands of documents highlight the need for grounding language models in reality, a role fulfilled by Retrieval-Augmented Generation (RAG). But here’s the catch: RAG isn’t one-size-fits-all.